ICMP-based m easurem ents (e.g. ping) are of ten criticized as un-representative of the applications’ experienced perform ance, as applications are based on TCP/UDP protoco ls and there is a w ell-accep ted conjecture that rou ters are often configured to treat ICMP differently from TCP and UDP. However, to the best of our knowledge, this assumption has not been validated. With this in m ind, we conducted extensive Internet end-to-end path m easurem ents of these three protocols, spanning ov er 90 sites (f rom both commercial and academ ic networks), over 6, 000 paths and m ore than 28 million prob es in PlanetLab during two weeks. Our results show that IC MP performa nce is a go od estim ator for TCP/UDP perform ance for the m ajority of the paths. However for nearly 0.5% of the paths, we found pe rsistent RTT differences between UDP and ICMP greater than 50% , while for TCP the difference ex ceeds 10% for 0.27% of the paths. Thus, although ICMP-based measurem ents can be trusted as predictors of TCP/UDP perform ance, distributed system s and ne twork researchers should be aware of some scenarios where th ese m easurem ents will be heavily m islead ing; th is paper also provides som e hints that can help in identifying those situations.